Including AI Prompts in Automations (Beta)

AI Prompts can be used to draft content, summarise documents and meeting records, extract key details or create structured output merge fields for later automation steps.

This article explains how AI prompts can be included in automations, how to review the AI conversation created when the automation runs, and what to check when testing an automation that uses AI.

Including Prompts in Automation Steps

AI Prompts can be added to the AI Prompt step or AI Tasks step as part of a custom process. The prompt tells AI what to do with the selected document, merge fields, or other context available to the automation.

For example, an AI prompt could review an incoming email and assess whether a reply is required, summarise a meeting record, extract details from a document, or generate keywords for smarter document filing. The AI Prompt step generates a Response and Output Merge Fields, which can be used as inputs or logic for later automation steps.

The AI Prompt Editor is where the prompt instructions are entered and configured. Depending on the step being used, the editor can include options for Web Search, FYI Merge Fields, and Output Merge Fields.

AI uses the information provided in the prompt, selected document, merge fields, and enabled tools such as Web Search. If Web Search is not enabled, FYI AI will not search the web. If the required information is not included, AI cannot reliably use it in the response.

Refer to Process Step Details - AI Prompt and Process Step Details - AI Tasks for detailed instructions.

Reviewing AI Conversations

When an automation runs that includes an AI Prompt or AI Tasks step, FYI creates an AI conversation that can be used to review what was sent to AI and what was returned.

The AI conversation is useful when testing or troubleshooting an automation. For example, if a merge field was not generated or a later step did not contain the expected value, review the conversation to confirm whether AI received the correct document and returned the expected Output Merge Fields.

To review an AI conversation:

1. Open the individual process and click the History tab.
2. Click the process in the list to open the Process Activity drawer.
3. Click the Conversation link to view the AI Prompt or AI tasks conversation.
   
4. The Conversation drawer will display.  From the drawer, review and confirm that the prompt included the correct instructions and context, and returned the expected Response and Output Merge Fields.

Conversation Drawer Details

Conversation Drawer Field	Details
Conversation	The original prompt and context, displayed at the top of the Conversation drawer. This includes a link to the document included for context, where applicable. Review the prompt and document to confirm the instruction was clear and included the required information. Tip: Click the arrow to expand or collapse the conversation.
Creating Merge Fields	The output merge fields created by the AI Prompt step, including the Response field, are displayed in a list below the AI conversation. Check the merge fields created by the AI Prompt step, including the default Response field. Tip: Click the arrow to expand or collapse the Response.
AI Tasks Created	When an automation with an AI Tasks step runs, the conversation will display: The original AI Tasks prompt sent to AI, displayed at the top of the drawer. This includes a link to the document included for context. List of created tasks. Summary of the AI Tasks response, displayed at the bottom of the conversation drawer.

Creating FYI AI Prompts

An AI Prompt is an instruction included within an AI Prompt or AI Tasks automation step. The prompt tells AI what response to generate and how the response should be structured.

Depending on the automation step, AI can use selected documents and merge fields as context. This allows the automation to generate output that is relevant to the document, client, user, or process where that context has been provided to the AI step.

A clear prompt helps FYI AI understand the task, the context, and the expected response. The prompt should explain what the AI should do, what information it should use, and how the response should be presented.

When writing a prompt, include:

• The task: Explain what the AI should create, summarise, identify, or extract.
• The context: Explain what information the AI should use, such as the selected document or merge fields. Add merge fields where the prompt needs dynamic information from FYI.
• The response format: Explain how the response should be structured, such as paragraphs, bullet points, task details, or a short summary.
• The rules: Explain what the AI should include, ignore, or do when information is missing. Avoid prompts that are vague or open-ended. For example, instead of asking AI to “summarise this”, explain the type of summary required, the audience, the length, and any details that must be included.
• Output Merge Fields: Use Output Merge Fields when later process steps need a specific value from the AI response. Each Output Merge Field can have its own individual instruction that tells AI exactly what value to return and the format to use

Example AI Prompts

The examples show how an AI Prompt can be structured for common automation scenarios. These examples should be adjusted to suit the practice's process, available context and required output.